Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences


    Ocean deoxygenation often takes place in proximity to zones of intense upwelling. Associated concerns about amplified ocean deoxygenation arise from an arguable likelihood that coastal upwelling systems in the world's oceans may further intensify as anthropogenic climate change proceeds. Comparative examples discussed include the uniquely intense seasonal Somali Current upwelling, the massive upwelling that occurs quasi-continuously off Namibia and the recently appearing and now annually recurring ‘dead zone’ off the US State of Oregon. The evident ‘transience’ in causal dynamics off Oregon is somewhat mirrored in an interannual-scale intermittence in eruptions of anaerobically formed noxious gases off Namibia. A mechanistic scheme draws the three examples towards a common context in which, in addition to the obvious but politically problematic remedy of actually reducing ‘greenhouse’ gas emissions, the potentially manageable abundance of strongly swimming, finely gill raker-meshed small pelagic fish emerges as a plausible regulating factor.

    This article is part of the themed issue ‘Ocean ventilation and deoxygenation in a warming world’.

    1. Introduction

    A fairly broad consensus is emerging that upwelling-favourable winds may generally intensify as greenhouse gas-mediated climate change proceeds [18]. Meanwhile, serious concerns are being expressed over a trend towards increased deoxygenation occurring within the world's oceans [911]. Since instances of deoxygenation are often observed in ocean regions that feature particularly intense upwelling circulations, identification and exploration of mechanisms that appear to dynamically link these two aspects would seem to be of major current scientific, as well as societal, interest.

    In the discussion presented here, three comparative examples are arrayed and scrutinized: (i) the Somali upwelling of the northwestern Indian Ocean, which during the height of the southwest monsoon becomes the strongest upwelling zone that exists anywhere in the world's oceans; (ii) the Namibian upwelling of the southeastern Atlantic Ocean which, by contrast, continues year round, while ranking as the strongest of the world's classical eastern ocean upwelling zones, and (iii) the Oregon dead zone that a dozen or so years ago abruptly appeared and has since become seasonally recurrent during the summer upwelling season off the US state of Oregon.

    2. Comparative examples

    (a) The Somali Current upwelling

    The strongest seasonal-scale wind-driven upwelling in the world's oceans (figure 1) occurs off the coast of Somalia during the boreal summer. The major driving force behind this exceptional upwelling system is the Findlater Jet, which stands as the strongest low-level atmospheric jet that exists as a regular feature anywhere in the world [14] and exists in reaction to the steep surface pressure gradient formed during boreal summer between the heated Asian land surface and the cooler surfaces of the Indian Ocean and Southern Hemisphere portions of the African landmass. This effect of land–ocean surface contrast is then strongly amplified by the strong convection caused by massive latent heat release as the rapidly accelerating near-surface air current, initially minimally retarded by coriolis constraints while traversing the near-equatorial zone is deflected upwards, first when encountering the coastal mountain ranges of the Indian subcontinent and then when forced to surmount the vertical barrier of the Himalayas. The net result is the uniquely powerful Findlater Jet air flow that extends from the coast of Somalia diagonally northwestward across the Arabian Sea. Within the coastal boundary upwelling zone off Somalia, this jet drives the strongest seasonal-scale wind-driven ocean upwelling that exists in the world's oceans [15] (figure 1).

    Figure 1.

    Figure 1. Comparison of characteristic seasonal maximum intensities of wind-driven offshore-directed surface Ekman transport at selected locations within the four major eastern ocean boundary upwelling regions of the world as well as along the boundary of the Arabian Sea. Units are metric tons per second transported offshore per meter of alongshore distance measured against a large-scale smoothed coastline trend. Indicated magnitudes are based on information contained in [12,13].

    The shear zones existing on both sides of the jet's core exert intense wind stress curl on the sea surface in a net cyclonic rotational sense along the northwestern flank of the jet and in a net anticyclonic sense on its southeastern flank [13]. This cross-jet contrast in wind-forced vorticity in turn drives a zone of very energetic ‘open ocean upwelling’ [16] on the left side of the jet (facing downwind) and a zone of comparably strong Ekman convergence and resulting intense downwelling on the right side, building up a narrow, very extended band of uniquely steep northwest to southeast upward slope in the sea surface that supports a geostrophic component of flow within the Somali Current, which at its seasonal maximum comstitutes by far the swiftest large-scale ocean current that exists on our planet [17].

    The conjunction of powerful offshore-directed wind-driven Ekman surface transport near the coastal boundary with the deeper underlying Somali Current flow acts to spew huge quantities of unoxidized organic production directly outwards into the Arabian Sea proper [15], where it eventually sinks and decomposes in the upper thermocline producing severe hypoxia at depths as shallow as 50–125 m over large areas of the northern Arabian Sea [18]. This zone exhibits the thickest low-oxygen layer to be found anywhere in the world's oceans [19] and ranks as one of the top three water-column deoxyfication sites in the world's oceans [20,21]; it has long been recognized as one of the world's most important zones of oceanic methane emissions [22,23]. During two full years at sea on the International Indian Ocean Expedition in the early 1960s performing at-sea analyses of nutrient chemistry [24], I personally experienced the acrid odour of hydrogen sulfide, both occasionally while drawing water from hydrographic samples taken from the upper thermocline or when bringing aboard putrid bottom trawl samples apparently entirely devoid of multicellular life; I also repeatedly encountered the very shallow intense spikes of dissolved nitrite that have been more recently reported by others [20].

    The intensely dispersive advection of all planktonic life appears to largely preclude effective grazing control by herbivorous zooplankton and the presence of significant populations of the pelagic-spawning small coastal pelagic fishes (sardines, anchovies, etc.) [15] that tend to dominate the pelagic biomasses and trophic flows in most ocean boundary ecosystems.

    The dissolved plant nutrients that are upwelled into the surface layers on the cyclonic side of the Finlater Jet are then carried across the underlying Somali Current in the Ekman transport field while supporting in its transit whatever degree of phytoplankton growth may be possible [25] under the intense turbulent mixing imposed by the overlying wind jet. This organic production is then forced downward and injected into the stratified ocean layers on the anticlonic side of the jet by the vigorous downwelling motion which is further augmented by passive sinking of cells due to the adaptively evolved negative bouyancy inherent to diatoms [26] and by even more rapidly sinking faecal pellets produced by whatever depth-maintaining herbivorous zooplankton that would accumulate [27] within, or just beneath, the convergent zones in the near-surface Ekman flow.

    While the intense turbulent mixing of the upper water column that undoubtedly acts to more or less constrain photosynthetic production [25] in the near vicinity of the Findlater Jet–Sommali Current complex, an additional class of phenomena that undoubtedly would act to enhance and distribute all the effects described above are the mesoscale and submesoscale eddy motions that are spawned by the immense mechanical energy transported by the coupled Findlater Jet–Sommali Current complex. The eddy kinetic energy existing in the Arabian Sea off the coast of Somalia has been estimated to be highest to be found anywhere in the world's oceans [28,29], acting to generate multiscale spatial and temporal arrangements of very strong upwelling and downwelling water motions [30]. The individual eddy structures themselves (figure 2) would continue to transfer their own momentum to adjacent water ‘parcels’, thereby spreading an expanding complex of spinning eddy vortices that would be subject to westward propagation via Rossby wave dynamics [27] and thereby act to expand the eddy distribution even counter to the advective directionality of the larger-scale air and water flows that had powered the development of the eddy motions in the first place. Thus, the eddy structures serve to transport eddy flow momentum and mechanical energy away from the zone of very strong turbulent mixing within and around the Findlater Jet–Sommali Current complex, to zones [25] where that energy can be more effectively used in powering organic production while deriving great advantage [30] from the interior organization and mechanics of the eddy structures themselves (figure 2).

    Figure 2.

    Figure 2. Schematic of pertinent physical–biological mechanisms acting within actively torqued ocean eddies. The illustrated rotational directions conform to a northern hemisphere context; for a Southern Hemisphere context, all aspects of the diagrams would be identical, except that the directions of rotation of the eddies would be in the opposite sense from that suggested in the drawing. (a) A cyclonic eddy (i.e. rotating counterclockwise in the Northern Hemisphere and clockwise in the Southern Hemisphere). (b) An anticyclonic eddy (i.e. rotating clockwise in the Northern Hemisphere and anticlockwise in the Southern Hemisphere). Note that eddies that would require less than a local half-pendulum day (i.e. less than approx. 1.5 days at 20° latitude, 2.9 days at 10° latitude, 5.7 days at 5° latitude, 28.6 days at 1° latitude, etc.) to complete one full rotation (i.e. Rossby number greater than 1) would be divergent within their interiors and so their patterns of interior flows and surface profiles would conform to diagram (a) rather than to diagram (b).

    For example, hydrostatic equilibration and fluid flow continuity would require that the associated sectors of sea surface divergence and convergence should be linked by connecting horizontal flows directed outwards from the divergent sectors towards interconnecting convergent sectors. The copious nutrients injected into the photic zone within the upwelling sectors would thereby be available to nourish a profuse proliferation of phytoplankton cells within a much milder turbulent mixing environment than would exist in near proximity to the Findlater Jet–Sommali Current complex. A follow-on planktonic trophic succession would in the meantime be passively transported towards advectively linked convergent zones distributed nearby, either in some different sectors of the same eddy (figure 2) or in an appropriate sectors of adjacent oppositely rotating eddies with which it may be exchanging momentum. Upon arrival, the entrained planktonic community that had been growing in abundance while in transit would be entrained in the associated strong downwelling motions (figure 2). Diatoms, due to their inherent negative bouyancy [26], would thereupon be subject not only to strong downward vertical advection but also to simple gravitational sinking. Zooplankton and other motile organisms that, even if unable to resist the more energetic horizontal flows, may be able, by their own swimming power, to maintain a preferred depth or light level within the much milder vertical flow field. These could, thus, be concentrated and maintained at a depth level (figure 2) where they could graze efficiently on the flow of diatoms and other organisms being vertically advected, as well as passively sinking, past them. In the process, these concentrations of depth-maintaining planktonic herbivores, omnivores and low trophic-level carnivores would be generating substantial amounts of even faster sinking faecal pellets to sink and decompose, either aerobically or anaerobically depending on the state of the in situ oxygen balances, fuelling deoxygenation processes in the stratified ocean layers just below. (Note that this same set of mechanisms would seem to offer quite a good fit also to the recently reported observations in the tropical North Atlantic Ocean of very near-surface dead zones associated with eddies emanating from the tropical West African upwelling zones [31].)

    (b) The ‘Lüderitz’ upwelling (southern Namibia)

    The strongest of the world's ‘classical’ eastern ocean boundary upwelling zones (figure 1) is located in the southwest Atlantic Ocean near Lüderitz, Namibia [32]. Whereas the Somali upwelling stands as by far the most seasonally intense of the world's regional coastal ocean upwelling systems, it remains so only for the several boreal summer months of the year during which the southwest monsoon holds sway. During the opposite northeast monsoon season, Somalia is actually a zone of rather intense wind-driven coastal downwelling [15]. By contrast, the Lüderitz system, although exhibiting a maximum seasonal upwelling intensity that may be less than 20% of that characterizing the seasonal monthly maximum off Somalia, continues to upwell subsurface waters nearly unceasingly throughout the year [15].

    Characteristics of ocean deoxygenation are starkly evident off Namibia. Surface signatures of massive eruptions of noxious gas products of anaerobic decomposition, some as large as the country of Wales (or the US state of New Jersey), have been frequently observed off the Nambian coast north (i.e. ‘downstream’) of the Lüderitz upwelling core [33]. Monitoring capability is enabled by highly reflective micro-granules of elemental sulfur that build up near the sea surface as a result of oxidation of upwardly erupted poisonous hydrogen sulfide (H2S) gas. These reflective particles produce a visible signature on the sea surface that is observable from earth-orbiting satellites. Throughout more than a year (2001–2002) of initial systematic observation, one or more of these major eruption signals were, more often than not, clearly visible somewhere along Namibia's coastline [33].

    During such eruptions, the ‘rotten egg’ smell of hydrogen sulfide permeates the local atmosphere. Associated highly corrosive sulfuric acid that is produced as erupted H2S gas becomes oxidized in the lower atmosphere and then dissolved in water droplets condensing in the coastal fog causes damaging corrosion of automobiles and other metal machinery and objects on the adjacent coastlands. A loss of two billion Cape hake (Merluccius capensis) was reported to have occurred during a single anoxic episode in 1991 [34], implying a loss due to this single event of a thousand large, highly palatable, readily marketable food fish for each man, woman and child living in Namibia, meanwhile episodically destroying 80% of the entire regional stock that had normally been the basis for Namibia's most valuable fishery [35]. Mortalities of near-shore fish and invertebrates have historically occurred with varying frequencies and intensities [36]; large accumulations of dead fish have often been left decomposing at the water's edge. Local residents, upon smelling the ‘rotten egg’ odour of erupted H2S gas, would sometimes rush to the ocean side to pick up rock lobsters (Jasus lalandii) that would, during eruptive episodes, be scrambling from the ocean onto the beach to escape the noxious conditions in the water [37].

    In these areas, dense populations of the world's largest bacterium Thiomargarita namibiensis employ nitrate ions gleaned from the hypoxic water column to oxidize H2S, thereby accessing energy to support their life processes [38]. Temporary floating mud islands have appeared on the sea surface comprising sheets of sea floor sediment that have become detached from the continental shelf surface and carried upward to the sea surface by the buoyancy of bubbles of H2S-laced methane (CH4) produced anaerobically at the sea floor and trapped within the pores of poorly consolidated sediments accumulated there [39]. In recent years, the formerly dominant sardine population has been largely supplanted by H2S-tolerant, hypoxia-acceptant pelagic gobies [40] and medusas [41,42].

    Puzzlingly, after nearly continuously appearing during the first year and a half of satellite monitoring, the eruptions suddenly ceased for the entire following year. At that time a brief notice appeared on the website of the Namibian National Marine Information and Research Center reporting that a ‘minor rebound’ in sardine reproductive activity seemed to be underway. This information prompted a simplified mathematical exercise [32] designed to explore the potential deoxygenation-triggering effects of various factors likely to be involved in the regulation of population densities of phytoplankton as well as of herbivorous zooplankton within the near-surface layers of such a highly productive, strongly divergent coastal upwelling zone. In this analysis, nutrients were not considered limiting with regard to photosynthesis within the near-coastal zone, although availability of light-absorbing chlorophyll (i.e. abundance of phytoplankton cells) remained a limiting commodity with respect to the overall rate of primary production.

    Equation (2.1) expresses on a ‘per unit volume’ basis the fractional rate of change of phytoplankton concentration (p) as an intrinsic growth rate under an excess of nutrients, from which are subtracted loss rates due to (i) offshore advection, (ii) deposition of sinking cells either onto the sea floor or into the stratified layers where both offshore Ekman flow and photosynthetic production essentially vanish, (iii) consumption by herbivorous zooplankton expressed as a product of a variable zooplankton concentration (z) that is, in turn, controlled by equation (2.2), and a constant consumption rate per unit concentration of zooplanktonic grazers, (iv) consumption by sardines expressed as a product of sardine concentration (biomass per unit volume) (S), and a constant consumption rate expressed as fraction of a unit volume of water filtered by a unit of sardine biomass per unit time, and (v) turbulent diffusion of phytoplankton particles either into or out of the zone. Equation (2.2) is a similarly formulated expression for the rate of change of concentration of herbivorous zooplankton, but without a ‘sinking/deposition loss’ term and with a ‘predation loss’ term that is not a function of other variables treated in the analysis.

    All of the factors linearly multiplying the variable p in equation (2.1) were assembled algebraically into a combined ‘α’ factor (expression (2.3)) and then substituted back into equation (2.1), yielding the much simpler-appearing equation (2.5). Similarly, the factors linearly multiplying the variable z in equation (2.2) were assembled algebraically into a combined ‘γ’ parameter (expression (2.4)) and then substituted back into equation (2.1), yielding equation (2.6).

    Display Formula
    Display Formula
    Display Formula
    Display Formula
    Display Formula
    Display Formula
    Display Formula
    Display Formula

    Equations (2.5) and (2.6) are easily integrated, and initial conditions applied to deal with the resulting constant of integration, yielding the respective ‘exponential response’ equations (2.7) and (2.8) which yield the not unexpected result that as long as, for example, the α parameter oscillates only very narrowly around a mean of zero, phytoplankton concentration remains more or less constant. However, if the value of α should fall into a durably negative value range, phytoplankton abundance declines exponentially, and if conversely α should durably rise into a positive value range, phytoplankton concentration increases exponentially. Essentially identical responses to the value, ranges of the ‘γ’ parameter operate with respect to zooplankton abundance. For more details on the mathematical manipulations, leading to these results, see [32].

    Schematic illustrations of the processes involved along with some implied effects are presented ‘cartoon fashion’ in figure 3. Figure 3a is meant to represent an upwelling system operating under a moderate level of forcing by fairly mild offshore-directed wind-driven surface Ekman transport, such as might characterize such coastal upwelling zones as exist off Oregon, Portugal, the Malabar Coast of India or northern Baja California (figure 1). In such situations of fairly modest rates of upwelling, planktonic organisms are not transported offshore so rapidly that the longer generation times of zooplankton compared with those of phytoplankton result in serious spatial separation of the respective patterns of abundance (although figure 3a is drawn to suggest initial rapid growth of phytoplankton abundance as nutrient-rich recently upwelled water moves into the illuminated photic zone in the left half of zone 2 and then begins to decline in the right half of zone 2 in response to lagged growth in the abundance of the more slowly responding zooplankton). Overall, grazing by herbivores acts to prevent uncontrolled phytoplankton proliferation in the nutrient-charged environment, while predation, as well as offshore advection, acts to keep the zooplankton abundance in check (i.e. during the upwelling seasons in such locations, the α and γ parameters would typically fluctuate in fairly narrow ranges around their zero levels). This predation by zooplanktivores sustains a local community of larger nektonic predators and forms a basis for viable local fisheries and other useful ecosystem services.

    Figure 3.

    Figure 3. Conceptual diagrams of important mechanisms, processes and outcomes operating within ‘temperate’ eastern ocean upwelling ecosystems (a) under moderately intense upwelling conditions, (b) under extremely intense upwelling conditions in which sardines are absent and (c) under extremely intense upwelling conditions in which sardines are regionally abundant.

    The ongoing wind-driven offshore movement of surface waters transports entrained planktonic organisms towards an offshore frontal zone (in zone 3 of the panel) where the more dense upwelling-conditioned coastal water tends to sink below less dense oceanic surface waters. Here, entrained organisms that are sufficiently motile are concentrated as they actively oppose being carried downwards beyond their preferred depth or light levels. The resulting concentrated zooplanktonic food source would attract small pelagic fishes and other planktivorous nekton, which would in turn attract larger pelagic predators and in turn attract fishing operations, all of which may use the flotsam and drifting objects that accumulate at the surface of the convergent frontal zone as a locational signal of the patchily distributed enriched prey concentrations likely to be found in the waters directly below [43].

    By contrast, figure 3b is intended to represent a much more intense coastal upwelling zone characterized by massive upwelling of dissolved plant nutrients into the photic layers within zones 1 and 2 of the panel. Details of the panel are drawn to reflect what is believed to have been, and continues to be, the characteristic marine ecosystem situation holding sway off the coast of southern Namibia for at least parts of the past several decades since the collapse of the formerly massive sardine fishery under heavy exploitation by unregulated foreign fleets.

    In zone 1, the intended impression is there may be very little ‘going on’ in a trophic sense within the water column. Planktonic organisms tend to be swept horizontally out of the upper layers of the zone by the intense offshore-directed flow. Photosynthetic production requires more than just nutrients and light; notably, chlorophyll to gather light energy needed to power photosynthesis. The strong offshore advection effectively opposes and thwarts shoreward turbulent diffusion of expelled cells back into the zone. Moreover, water being upwelled to the surface from the light deficient layers below would contain few viable, light-adapted, photosynthetically ready phytoplankton cells. And so, until there are enough cells accumulating, growing, dividing and thereby proliferating rapidly enough to stem the effect of offshore advection of everything that may have been formed, the building up of a primary food base for an in situ multi-levelled trophic web is a slowly developing process. Accordingly, the zone would appear to offer little to attract fish or other nekton possessing sufficient swimming capability to effectively counter expulsion by the strong offshore advection. Indeed, actual data identify the zone near Lüderitz as representing an effective barrier to migratory exchanges of sardines and other fishes [44].

    But, as suggested in figure 3b, as the flow proceeds from zone 1 to zone 2, a growing abundance of phytoplankton cells would be expected to gradually develop and zooplankton abundance to begin to grow in concert. However, while a phytoplankton cell may reproduce and multiply by a simple process of cell division, a copepod needs to grow through five to six naupliar stages and a similar number of copepodid stages in order to reach reproductive maturity [45]. As a result, the zooplankton population growth rate (r(z)growth) within the composite γ parameter tends to be slower than the phytoplankton growth rate (r(p)growth) within the corresponding α parameter. And the ‘offshore advective loss' items, that in such an intense offshore-directed flow situation are dominant terms, retains the same ‘per unit biomass’ magnitude in both the ‘α’ and ‘γ’ cases, because it depends solely on the rate of volume transport of the water in which both populations are assumed to be passively entrained.

    Meanwhile, the zooplankton (z) and phytoplankton (p) components are interacting via the ‘grazing loss rate’ item (zcgrazing) in the α parameter. This ‘grazing rate’ item, because it is formulated ‘per unit phytoplankton concentration’, decreases in absolute magnitude as phytoplankton biomass increases, while increasing in absolute magnitude as zooplankton biomass increases. Since zooplankton biomass, because of its slower population growth rate, would be increasing more slowly than phytoplankton biomass, at some point (illustrated near the left side of zone 2), the declining absolute magnitude of the negatively signed ‘grazing loss rate’ rapidly turns the value of the α parameter increasingly positive. This represents a sharply accelerating disruption of effective grazing control, and phytoplankton abundance thus grows exponentially according to equation (2.7) within this grazing ‘loophole’. A profusion of sinking diatom cells rains down, forming a thick carpet of incompletely oxidized organic matter that covers the floor of the continental shelf, the interior layers of which may quickly turn anoxic as the available dissolved oxygen is used by proliferating bacteria. When oxygen is totally spent, the decomposition turns anaerobic, generating the noxious gaseous products that build up to the point that the combined buoyancy of the myriads of expanding bubbles of methane gas may overcome the restraints of surface tension and adhesion and abruptly initiate the gaseous eruptions described earlier in this section. In the process, the water column may be essentially ‘sterilized’ by the cloud of bubble-borne poisonous, H2S-infused methane (CH4) and other gases rising through it. Moreover, even after the noxious and poisonous gases have dissipated, a more durable effect is the lingering deoxygenation that occurs as the H2S in the micro-bubbles of methane rising through the increasingly oxygenated water nearer the sea surface is progressively oxidized to the highly reflective elemental sulfur micro-particles that form the visible sea surface ‘signatures’.

    In the meantime, zooplankton and other organisms not mortally impacted would continue to be advected offshore where those possessing adequate motility to maintain a preferred depth/light level will be accumulating in the convergent frontal system that is indicated by the blue-dashed curved line near the left side of the panel. This accumulation would constitute a concentrated food source for zooplanktivores, including sulfide-tolerant, hypoxia-resistant pelagic gobies [40] and medusas [41,42], forming a ‘critical mass’ for these organisms to proliferate, spread out and ultimately infest much of the overall system, including the favourable spawning habitats of important local fish stocks [44].

    Certainly, fish eggs and newly hatched fish larvae would seem to represent ideal prey for medusas, pelagic gobies and other zooplanktivorous organisms. But, fish eggs and fish larvae alone would normally be insufficient to sustain burgeoning zooplanktivore populations, which would also need to forage on much more abundant food organisms such as herbivorous copepods. As the infestation of zooplantivores grew, the negatively signed ‘loss rate of herbivorous zooplankton due to predation’ item (rpredation) would steadily increase in magnitude, eventually turning the γ parameter itself negative, leading to exponential decline in herbivorous zooplankton abundance, and in consequence, to exponential increase in phytoplankton abundance. This would work to ‘lock’ the system in the unfortunate state depicted in figure 3, panel 2. The chain of causative events suggested here could perhaps be considered as an extreme example of a backwards-folded ecosystem response curve, as set forth in the widely quoted essay ‘catastrophic shifts in ecosystems' by Scheffer et al. [46].

    Figure 3c depicts quite a different potential outcome within the same extremely intense Namibian coastal upwelling system configuration. But, in this case, one must imagine that the 10 million metric ton sardine population [47] that was eradicated some four decades had somehow been re-established as a dominant component of the regional ecosystem. Sardines are a noteworthy species group among the important fish species of upwelling regions in their very fine-meshed gill raker structures that allow them to effectively filter and directly consume microscopic phytoplankton [48], in addition to also consuming larger zooplankton by either filtering or by raptorial feeding. This renders them facultative omnivores. Moreover, they are very strong swimmers and prodigious migrators, and so, in contrast with much weaker-swimming herbivorous zooplankton, are easily capable of countering the strong offshore-directed surface flow near the upwelling zone to directly access the phytoplankton concentrations as well as whatever zooplankton accumulations that may be entrained with them as the planktonic mixture is transported offshore. Sardines seem also to be exceptionally opportunistically inclined [45] and habitually move in very large aggregations such that, in periods of high biomass, individual schools may commonly contain more than 2 million fish, while a significant fraction of the schools may contain up to 10 million or more [49]. When many schools group together in massive migrating shoals, such as in the famous KwaZulu-Natal sardine run [50], the amalgamated shoals have been reported to stretch up to 15 km in length, 3.5 km in width and 40 m or more in vertical thickness. Clearly, an incursion of grouped actively feeding sardine schools could be capable of radically reducing the concentration of photosynthesizing diatom cells within some segment of the highly divergent coastal upwelling zone (e.g. zones 1 and 2 depicted in figure 3). And once the local diatom abundance has been essentially obliterated, and the sardines schools have, in consequence, moved on to exploit a fresh habitat segment, the phytoplankton biomass must be slow to replenish itself in the resulting situation of minimal chlorophyll available to absorb light energy, of particularly strong, continuing advective loss that effectively counters any influx by lateral diffusion and of the paucity of photosynthesis-ready phytoplankton cells that may be injected due to upwelling from below the photic zone.

    Accordingly, if regional sardine abundance could be restored to a level approaching that existing prior to their 1970s collapse, a situation where (according to the formulation presented in [32] and briefly outlined in this section) both γ and α would tend to slip into near-zero value ranges maintained by the opportunistic feeding and migratory behaviours of schools of omnivorous sardines. This could potentially bring on a series of sweeping changes whereby abundances of planktonic organisms would be much reduced and the downward rain of incompletely oxidized diatom cells and other organic matter could essentially cease, relieving the other undesirable features depicted in figure 3b. However, the copious nutrients supplied by the intense upwelling would not be lost, but would only cease being deleteriously trapped within the very near-coastal zone, remaining in the surface layers to, as those layers are transported offshore, be available for supporting a more desirably configured trophic system in the wider ecosystem at large.

    As planktonic food concentrations declined in the primary upwelling zone, opportunistic sardines could assumedly move to better feeding grounds, such as the offshore front depicted in figure 3, zone 3, depriving the obligate zooplanktivores (medusas, pelagic gobies, etc.) of food concentrations sufficient to maintain their critical masses, and also probably, in at least some cases, preying on early life stages of the medusas and perhaps other zooplantivorous species that may be attempting to reproduce there [51]. Thus, the infestation of the regional system by predators on planktonic early life stages of more desirable species might be reversed, allowing stocks of valuable food fish and invertebrates to perhaps rebuild.

    (c) The Oregon ‘dead zone’

    About a dozen years ago, a new hypoxic dead zone abruptly appeared in the northeastern Pacific Ocean off the coast of Oregon [52,53]. Although, it had never been reported earlier, the dead zone phenomena at that location have since become reliably recurrent during the summers of subsequent years. Piles of dead fish lining the waters edge very much mirror those on the beaches of Namibia following anoxic episodes. Also, the abrupt appearance and continuing summer recurrence of the Oregon dead zone phenomena suggests a similar tendency towards a ‘abruptly triggerable regime shift’ as does the sudden total cessation of H2S eruptions off Namibia. However, one notable conundrum needing rationalization is that contrary to the extremely intense cases of Namibia and Somalia, the wind-induced upwelling and associated offshore advection off Oregon is comparatively weak (figure 1).

    Charles Darwin famously said that ‘without speculation there can be no true or original observation’. Following this advice from our greatest scientific hero, I would like to speculatively pose the idea of a coastal-trapped wave [54] connection to the dead zone location off northwestern Oregon from the much stronger upwelling that abruptly begins in southern Oregon [55] near (Cape Blanco) a relatively short distance (approx. 200 km) to the south and grows to the even stronger upwelling intensity existing in northern California near Cape Mendocino (figure 1), an additional (approx. 200 km) distance further south.

    For a simplified explanation that might succeed in conveying at least the gist of the argument, consider figure 4, in which panel t0 is intended to represent the upwelling region extending from northern California northward nearly to the Canadian border, in which intense summer seasonal heating of the coastal landmass in the southern part of the region produces a cross-coastal atmospheric pressure gradient, which supports a balancing northerly (southward-directed) geostrophic wind that in turn exerts a southward frictional stress on the sea surface, driving an offshore-directed Ekman transport of ocean surface water.

    Figure 4.

    Figure 4. Sequential diagrams of the manner and mechanisms by which the effect of local wind-driven offshore-directed Ekman transport may propagate poleward along an eastern ocean continental boundary to enhance locally driven upwelling along its path, while in the process sometimes also generating deoxygenation and other undesirable ‘side effects’. (Symbols shown connote the same meanings as do the corresponding symbols appearing in figures 2 and 3.)

    Panel t1 shows the slight depression in the sea surface topography (i.e. the sea level ‘low’, symbolized in the diagrams by the capital ‘L’ symbols) produced by the transport of ocean surface water away from the solid coastal boundary and the geostrophic current (represented by the ‘arrow’ symbols) that will have developed around the sea surface ‘low’ on a time scale of a half-pendulum day (approx. 18 h at this latitude). Since the water that would be being carried by the current would have to come from somewhere, there would be flow divergence just beyond the north end of the ‘low’, acting to lower the sea surface level at that location. Accordingly, the depression would tend to propagate its northern edge ever further northward (panels t2 and t3). Also being illustrated are the increased geostrophic current velocities in response to the increased transverse pressure gradient due to the propagation of the sea surface ‘low’, and the tendency for the current, once it is beginning to run out of transverse pressure gradient due to the northward propagation, to broaden and veer offshore to contribute its transported water to the offshore-directed surface Ekman flow. Panel t4 indicates the continuing strengthening and narrowing of the current flow along the isopleths of steepening topographic gradient to form a continuous, linked coastal upwelling jet/coastal upwelling front structure [56]. Panel ‘t5’ displays the symbolic indications (employing the same symbols as appear in figures 2 and 3) of the nutrients, phytoplankton and zooplankton that would be present and subject to passive horizontal advection within the narrow upwelling front/jet current zone.

    Importantly, the jet current is a narrowed flow that is imbedded in the narrow, sharply sloping upwelling frontal zone which responds to, and must maintain flow continuity with, the broader (and therefore slower in actual flow speed) offshore Ekman transport distribution. Accordingly, the actual advective flow speeds within a more voluminous but much broader offshore-directed Ekman transport field (e.g. off Lüderitz, Namibia) and the advective flow speeds within the narrow upwelling jet current (off Oregon) might be more comparable than a straight comparison of local Ekman transport magnitudes (figure 1) might initially suggest.

    Figure 4b suggests a sequence of responses to climate change that could reasonably be expected based on inferences that emerge from the two earlier examples featured in §2a,b of this essay. Panel t6 calls attention to climate change-related intensification of the coastal thermal ‘low’ and the resulting non-linear (i.e. approximately quadratic) increases in upwelling-favourable wind stress and associated offshore-directed upper layer Ekman transport, which leads to increased downstream acceleration of the upwelling jet current (symbolized by the lengthened arrow symbols in panel t7) and of the contiguous advection within the upwelling jet/front complex and directly adjacent southward flowing waters. Panels t8t10 sequentially illustrate the progressive spatial separation of the distributions of the less rapidly lifecycling zooplankton from those of the faster-cycling phytoplankton. This allows a ‘loophole’ to open in the grazing control exerted on phytoplankton growth abruptly turning the ‘α’ parameter more and more positive. As a result, the phytoplankton biomass within the ‘loophole’ zone would be released to grow without limit, triggering a downward ‘rain’ of incompletely oxidized diatom cells into the stratified portion of the water column, or onto the surface of the continental shelf, where they could exhaust in situ supplies of dissolved oxygen (as symbolized by the black ‘X’ symbols that are shown superimposed on the ‘loophole’ zone in figure 4, panel t11).

    One might note here, in view of the suggestion in §2b, that a collapse in regional sardine abundance might have been a contributing factor in the development of the degraded situation off Namibia that sardine abundance in the northeastern Pacific region appears to have been reduced by an estimated 91% within the past decade [57].

    Finally, widespread hypoxic and anoxic conditions have been also reported [58] in the upwelling region that stretches along the coast of western India that forms the eastern boundary of the Arabian Sea, a region that encompasses the Malabar Coast where, similarly to the Oregon situation, figure 1 shows more mild locally wind-driven offshore Ekman transport compared with a much stronger offshore Ekman transport that apparently exists near Cape Cormoran at the southern extremity of the Indian subcontinent. One may wonder therefore whether an analogous coastal-trapped wave-linked mechanism to that suggested (figure 4) for the Oregon case might also be operating off western India. A quick perusal of figure 1 will perhaps also suggest some other instances (e.g. San Juan southward, Cap Blanc northward) where similarly configured coastal-trapped wave connections between zones of greater wind-driven offshore Ekman transport might serve to significantly enhance the upwelling occurring within a zone of less intense offshore Ekman transport located poleward along an eastern ocean boundary (or equatorward along a western ocean boundary), as well as generate unexpectedly rapid advective velocities within the ‘frontal jet’ current imbedded within the narrow coastal upwelling frontal zone that demarcates the upwelling-conditioned coastal water from the less dense oceanic surface waters existing further offshore.

    3. Discussion and concluding comments

    Each of the three examples examined have yielded similar, although slightly differently configured, sets of shared conceptual elements. The Somali and Namibian examples, while existing in very different geoclimatic circumstances, feature the two most intense regional-scale ocean upwelling cells that exist on our planet, and so their features that may be unique compared with those of the weaker upwelling systems (figure 1) may signal changes that may be expected to become more general and ever more extreme as climate change acts to augment land–sea temperature contrasts. The discussion of the rather ‘mild’ local upwelling situation of the Oregon example and the proposed coastal-trapped wave linkage mechanisms that brought it more into the ‘intense advection’ context of the earlier two examples, while providing evidence of a similar ‘triggerable’ system instability in the abrupt unprecedented appearance of the ‘dead zone’ that in the ensuing years became a rather consistent seasonally recurrent feature [53,54].

    To briefly recapitulate, the intense flow situations involved in both the Somali and Namibian examples, as well as speculatively associated with the coastal-trapped wave-linked mechanism proposed for the Oregon example, could tend to produce such large ‘advective loss’ terms in expressions 1 and 2 of §2b that they may, by themselves, come near to balancing (i.e. numerically offsetting) the respective intrinsic growth terms. Thus, even rather gradual trends in flow velocities, or in growth or decline in the presence of particular agents of consumption (e.g. strongly nektonic herbivores/omnivores such as sardines), could bring the respective dynamic balances (represented in equations (2.3) and (2.4)) inexorably towards ‘tipping points’ where a respective α (or γ) parameter could abruptly ‘tip over’ into an ‘amplifying’ positive (or ‘collapsing’ negative) range (equations (2.7) and (2.8)) and foment a radical shift in ‘regime’. The intrinsic growth rate of phytoplankton, which reproduces by simple cell division, would tend to be more rapid than that of herbivorous copepods, which generally must pass through five naupliar stages and five or six copepedid stages to become reproductively mature. Accordingly, because the advective loss rates would tend to remain similar in both, the tipping point of the γ parameter to negative ranges would likely be situated earlier along the advective path than would be that of the α parameter, opening an intervening ‘loophole’ within which control exerted by zooplanktonic grazing pressure would disappear. This, in turn, unless large numbers of opportunistic sardines were regionally available to migrate en mass to take advantage of the sudden hugely enhanced food availability in the loophole zone and thereby effectively re-establish grazing control, could tip the α parameter into a strongly positive range, setting loose uncontrolled growth of phytoplankton abundance and the follow-on effects discussed in §2b and illustrated in figure 2b.

    For example, the decimation of regional sardine abundance off Namibia by massive overfishing by foreign fleets in the 1970s, which can be hypothetically viewed as having tipped the regional-scale γ parameter (expression (2.2), above) to negative values thereby tipping the regional α parameter (expression (2.1)) into a positive value range, initiating (as well as reflecting) a tangled complex of interlinked fishery–ecosystem feedbacks [51] that over the immediately following decade or two seem to have acted to durably switch the situation from one initially resembling the depiction in figure 3c to one resembling that represented in figure 3b. In the case of Oregon, a gradual climate change-associated accelerating trend of the flow within the upwelling jet current (figure 4), perhaps abetted by a decadal-scale decline in sardine abundance [57], could finally have tipped the local γ parameter into a negative value range, abruptly opening the developing ‘loophole’ in herbivory depicted in figure 4, panels t8t10, thereby leading finally to the deoxygenation and ‘dead zone’ formation (represented in panel t11) within that ‘loophole’ area. In the Arabian Sea case, the extreme swiftness of the Somali Current flow may act to keep both α and γ constrained within the negative value ranges within the current itself, while the huge momentum stored in the massive flows fuels the ‘spin up’ of vigorous eddy fields [28] that produce, via mechanisms of the sort diagrammed in figure 1, the widespread, somewhat patchy, near-surface deoxygenation effects discussed in §2a. Importantly, all three examples along with their inferred mechanisms lead to the same overall conclusion, i.e. that in all of the three cases examined, climate change will most probably act so as to contribute towards increasing the overall rates of ocean deoxygenation.

    An additional speculative but interesting inference that has emerged from this brief examination of these three example regional upwelling systems was an implied beneficial impact of keeping sardine-type (omnivorous, strongly swimming, finely meshed-filter-feeding) nekton operating as major components of these types of ecosystems. This would seem likely to be vanishingly inexpensive in comparison to actually dealing with the ‘greenhouse gas’ side of the problem. And even, if it would be merely ‘putting a BandAid on a cancer’, working to maintain the evolved structure of a natural ecosystem would seem almost never to be a very bad idea. Obviously, one would hope that conscientious scientists would be using their training and talents to productively work all sides of this interacting mixture of science, economics and politics.

    Data accessibility

    This article has no additional data.

    Competing interests

    I declare I have no competing interests.


    I received no funding for this study.


    One contribution of 11 to a discussion meeting issue ‘Ocean ventilation and deoxygenation in a warming world’.

    Published by the Royal Society. All rights reserved.